1. art and entertainment
2. celebrity fan and gossip

Celebrity Endorsements, Firm Value and Reputation Risk:
Evidence from the Tiger Woods Scandal
Christopher R. Knittel and Victor Stango∗
February 9, 2012
∗
Thanks to Anson Soderbery for fast and thorough research assistance. Knittel: William Barton Rogers Professor of Energy Economics, Sloan School of Management, Massachusetts Institute of Technology and NBER, email:
knittel@mit.edu. Stango: Graduate School of Management, University of California, One Shields Avenue, Davis CA
95616. Email: vstango@ucdavis.edu.
Celebrity Endorsements, Firm Value and Reputation Risk:
Evidence from the Tiger Woods Scandal
Abstract
We estimate the stock market effects of the Tiger Woods scandal on his sponsors and sponsors’ competitors. In the 10-15 trading days after the onset of the scandal the full portfolio of
sponsors lost more than two percent of market value, with losses concentrated among the core
three sponsors EA, Nike and PepsiCo (Gatorade). Sponsors’ day-by-day losses correlate strongly
with Google search intensity regarding the endorsement-related impact of the scandal, as well as
with qualitative indicators of “endorsement-related news.” At least some sponsors’ losses were
competitors’ gains, suggesting that endorsement deals are partially a business-stealing strategy. However, competitors who were themselves celebrity endorsement-intensive fared relatively
worse than those who were not endorsement-intensive, and that difference also correlates dayby-day with news/search intensity regarding the scandal. It appears that the scandal sent a
negative market-wide signal about the reputation risk associated with celebrity endorsements.
Keywords: celebrity endorsers, event studies, reputation risk
1
Introduction
As of mid-2009 professional golfer Eldrick ‘Tiger’ Woods earned roughly $100 million annually in
endorsement income, an amount far greater than that earned by any other athlete. On November
27, 2009, Woods was involved in a car accident outside his home. Following the accident, a series
of news reports about both the crash and Woods’ personal life damaged his public reputation, and
several sponsors either stopped featuring him or dropped him outright. In this paper we estimate
the stock market effects of the scandal, for both the sponsor firms and their competitors. Some of
those competitors are themselves “endorsement-intensive” (but have no deal with Tiger Woods),
while others have no celebrity endorsement deals.
Our empirics address several key questions about celebrity endorsements, firm value and business strategy. Does firm value depend materially on investments in celebrity endorsements? If
so, do sponsors’ gains and losses from celebrity endorsements represent net market value creation/destruction, or business-stealing from other firms? And, does the stock market reflect changing expectations about the “reputation risk” that firms take on by attaching their brands to celebrities? Previous work on celebrity sponsorship almost exclusively focuses on the first question, rather
than the latter two. And even the work on gains from sponsorship faces some econometric difficulties that we circumvent, by dint of examining the downside of a scandal rather than the upside
of the initial endorsement deal. We also employ novel auxiliary data from Google Insights that
allow us to correlate endorsement-related news/search intensity with changes in firm value; to our
knowledge, ours is the first paper to use internet search intensity to understand changes in brand
value.
Our first empirical finding is that between the car accident and Woods’ announcement ten trading days later of an ‘indefinite leave’ from golf, his sponsors’ overall market value declined by over
two percentage points. This holds whether we measure losses relative to the stock market overall,
or relative to both the overall market and competitor firms in the sponsors’ primary industries.
Narrower groups of “Primary” firms with the biggest endorsement contracts, or that had made
large complementary investments in the “Tiger brand,” lost more in percentage terms. The losses
grow further by fifteen trading days after the accident.
We sharpen the empirics by showing a strong relationship between daily abnormal returns and
several measures of endorsement-related news/search intensity during the scandal. For example,
during the scandal sponsors’ losses are greater on days when the search term “Tiger Woods en-
1
dorsement” is more popular on Google, a result that is statistically significant and economically
substantive. For Woods’ core three “Tiger Brand” sponsors Google search intensity explains over
thirty percent of variation in abnormal returns during the fifteen trading days after the onset of
the scandal; the figure is lower but still significant for the full set of sponsors. The quantitative
search intensity outperforms an author-defined variable denoting significant “endorsement-related
news days.”
We also estimate stock price changes for sponsors’ competitors. We find that as sponsors lost
market value, competitors gained market value, as long as those competitors were themselves not
heavily invested in celebrity endorsements. Sponsors’ competitors with at least one celebrity endorsement deal experienced returns that are statistically significantly smaller than those experienced
by competitors without any celebrity endorsement deals, and close to zero on net. The day-to-day
pattern of competitors’ abnormal returns correlates strongly with both sponsors’ returns, and with
our auxiliary measures of news/search intensity; on days of high search interest in the term “Tiger
Woods endorsement,” non-endorsement-intensive competitors’ gains are more positive, and more
positive relative to endorsement-intensive competitors.
In the context of prior work linking stock market value to celebrity endorsements, our first
result provides clear evidence that in this case, a celebrity endorsement substantively affected stock
market value for sponsor firms. The losses that we measure are the converse of stock increases
one would suppose were generated initially through Tiger Woods’ endorsement deals. Previous
evidence of links between endorsements and stock market value has been mixed, because nearly
all of that work faces a harder identification problem: it uses initial endorsement announcements,
which are likely to be at least partially anticipated by traders, to estimate gains in firm value.1 The
event we examine was by all accounts a complete surprise to the market, making it a near-ideal
natural experiment from an event study perspective.
A corollary of our result is that endorsement deals carry substantial risk. While we cannot
compare the losses sustained by sponsors to their initial gains, the losses we estimate are large.
That suggests taking a view of celebrity endorsement as a risky investment rather than a simple
short-run cost-benefit tradeoff—particularly if a firm plans to complement the endorsement deal
with co-investment in a new product or brand, as Nike did with its golf line, and as Electronic Arts
and Gatorade did with their “Tiger-specific” products.
Our finding that sponsors’ losses are competitors’ gains is fairly novel in the context of previous
1
Louie, Kulik, and Jacobson (2001) is a notable exception. We discuss that work below.
2
work correlating endorsements with firm value. We are aware of one previous study (Mathur,
Mathur, and Rangan (1997)) examining competitors’ returns after Michael Jordan’s announced
return to professional basketball, but that study finds “only very weak evidence” of a link between
an endorser’s behavior and competitors’ stock market value. For business strategy, the upshot of
our finding is that one could view celebrity endorsements as yet another tool for stealing business
from competitors.
Important corroborative evidence for these findings, albeit using a completely different method
and data set, comes from a recent paper by Chung, Derdenger, and Srinivasan (2011). That paper
estimates a structural demand model of the golf ball industry, and uses the Tiger Woods scandal to
identify changes in demand. The authors find that demand for Nike golf balls shifts down following
the scandal, significantly reducing Nike’s flow of profits from selling golf balls. The empirics suggest
both that total demand for golf balls fell (i.e., that there is a category effect), and that competitors
of Nike experienced relative gains (i.e., that there is a business-stealing effect).
We view our incorporation of Google Insights search intensity into the empirics as particularly
promising for future work in marketing. A small but rapidly growing set of papers in finance establishes that Google search intensity is correlated with stock prices more generally (see in particular
Da, Engelberg, and Gao (2011) and papers citing that work). But we are aware of no other work
showing that search intensity in marketing-related domains like celebrity endorsements also has
significant power to explain stock price changes.
Finally, the difference in competitors’ returns when we stratify by competitors’ “endorsementintensity” is provocative evidence about how markets price reputation risk associated with celebrity
endorsements, and about how events can change perceptions of that risk. The relatively more
negative returns for endorsement-intensive competitors suggests that the scandal changed marketwide perceptions of risk associated with investments in celebrity endorsement. We are not aware
of any previous work examining this issue, and in the conclusion we discuss the implications of this
finding in more detail.
2
Celebrity Endorsements and Firm Stock Market Value
Celebrity product endorsements, and endorsements by professional athletes in particular, are a
critical element of brand strategy.2 The key question from a firm’s perspective, of course, is
2
See, e.g., the many references in Ding, Molchanov, and Stork (2008), and an earlier survey by Erdogan (1999).
3
whether a celebrity endorsement generates value sufficient to offset its possibly considerable cost.
Quantifying that benefit-cost tradeoff is hard, and consequently the question of whether celebrity
endorsements are value-enhancing remains open.
Stock market event studies provide one window into measuring the returns associated with
celebrity endorsements. A firm’s stock price reflects expectations about the discounted value of
future economic profits. If retaining a valuable endorser changes those expectations—say, by increasing expected future sales—then an announcement of celebrity endorsement should generate a
“kick” in the stock price. Conversely, an adverse (reputation-damaging) event or the departure of
a valuable endorser might move those expectations about future profits downward, which should
result in a lower stock price.
Another dimension of using stock prices to evaluate celebrity endorsements is risk. As with
any investment, there is a chance that an endorsement deal will not pay off, either because a firm
initially underestimates the true gain associated with endorsement, or because the added value
of the celebrity endorser falls. Investors should treat that “reputation risk” as they would treat
any other component of risk in a firm’s stock: higher risk is less attractive. Holding the expected
level of future profits constant, investors should punish riskier firms with lower stock prices. In the
context of celebrity endorsements, that means that any firm with substantive exposure to celebrity
risk should be priced accordingly. More important, it means that changes in how markets perceive
the risk of celebrity endorsements might affect the value of all firms with celebrity reputation risk
exposure.
Beyond those straightforward intuitions, there is nuance to the stock market-based method
of measuring returns from endorsements. Stock prices reflect changes in expected profit rather
than sales or market share. Given that endorsement incurs expenses, it is possible that a celebrity
endorsement might reduce profit even as it sparks sales or growth. Put more formally, celebrity
endorsements generate economic rents, and the terms of the endorsement deal divide those rents. It
is possible that celebrities might bargain away all of the rents that they generate for their sponsors,
making sponsorship at best a break-even proposition. On the other hand, higher stock market
prices for sponsors indicate that the firm has captured some of the economic rents generated by the
endorser/firm partnership. The key question for a firm, then, is whether it is possible (or perhaps
likely on average) that firms can capture rents generated by celebrity endorsements.
Another point worth mentioning is that because changes in expectations drive changes in stock
prices, it is much harder to measure changes in firm value following well-anticipated events. If, for
4
example, a celebrity endorsement deal is widely anticipated long before its formal announcement,
buyers and sellers of the sponsor’s stock will have fully priced all of the gains associated with
the deal well before the announcement itself, and the actual announcement will change neither
expectations nor stock prices. Examining stock price movements around the actual announcement
could therefore understate the gains associated with the endorsement deal. That means that the
empirically cleanest type of event to use for quantifying changes in firm value is a surprise, whether
it is good or bad, because surprises by definition avoid the anticipation problem.
In the context of the identification issue on the front end, it is not surprising that previous
studies attempting to link celebrity endorsements and corporate sponsorship to stock market value
have found mixed evidence.3 We are aware of one study examining announcements of “bad news”
for celebrity endorsers (including athletes and entertainers); bad news is often, though not always,
more of a surprise than announcements of endorsement/sponsorship deals, and therefore provides
cleaner identification. In that paper, Louie, Kulik, and Jacobson (2001) find that bad news with
little “culpability” for the endorser (such as a career-ending injury) generates gains for sponsors—
this is an “any publicity is good publicity” result—while bad news with more culpability (such as
a DUI arrest) generates losses.4 The scandal that we examine falls squarely in the second (“more
culpability”) class.
Previous studies also may contain mixed findings for two other reasons. First, it is probably
true that while some firms may capture rents when they sign celebrity endorsers, others may not.
Some celebrities may command payments that completely offset any incremental profit generated
for the sponsor firm. And second, some firms may simply overestimate the gains associated with
an endorsement deal; by a winner’s curse logic, those firms should in fact be the ones who sign
celebrities more often.
One advantage in our case is that the scandal represented a surprise. Before the accident, Tiger
Woods was widely acknowledged to have the most valuable “brand” of any athlete in the world—a
fact accruing both from his athletic success and from his clean public image. Until 2009 he routinely
3
Farrell, Karels, and Montfort (2000) find that Tiger Woods’ endorsement deal announcements generated stock
market value for Nike, but not for American Express or Fortune (Titleist). Agrawal and Kamakura (1995), Mishra, Jr,
and Bhabra (1997), Miyazaki and Morgan (2001), Pruitt, Cornwell, and Clark (2004) and Samitasa and Kenourgiosb
(2008) find that endorsements/sponsorships generate positive stock market returns. Mathur, Mathur, and Rangan
(1997) find that Michael Jordan’s return to professional basketball generated positive returns for his sponsors. find
that celebrity endorsements generate positive stock market returns for a wide set of celebrities. On the other hand,
Fizel, McNeil, and Smaby (2008), Farrell and Frame (1997), Clark, Cornwell, and Pruitt (2009), Cornwell, Pruitt, and
Ness (2001) and Ding, Molchanov, and Stork (2008) find weaker evidence, or even evidence (in the case of Olympic
sponsorships) negative returns following endorsement/sponsorship announcements.
4
That paper also adds to an interesting set of studies asking how negative information about an endorser affects
brand perception and firm value. See, e.g., Till and Shimp (1998).
5
placed in the top 5 of the Forbes “Celebrity 100” list of most influential celebrities world-wide. So
our setting is certainly one in which stock prices might plausibly reveal the economic object of
interest, because there is no evidence that the market anticipated any of the bad news associated
with the scandal. The flipside of that, and a limitation of our approach, is that while our method
can estimate by how much sponsors’ expected future profits fall after the scandal, it cannot estimate
the gain in expected future profits that firms initially experienced from the endorsement deal.
Another benefit associated with our example is that Tiger Woods endorses several products
rather than just one. This allows us to estimate stock market effects across a wide set of otherwise
unrelated firms, and gives us more statistical power than one would have if the estimates were
confined to a single sponsor firm.5 Comparing returns for many sponsors associated with a single
endorser can shed light on the circumstances in which endorsement deals are profitable for firms,
as long as one properly controls for the contemporaneous correlation in errors across sponsor firms.
We can further improve the power of our tests by exploiting information about how returns
and information co-move (or do not co-move) during the time period of the scandal. Although the
scandal was a surprise, news related to the scandal, and endorsement-related news in particular,
disseminated gradually after the date of the accident, and did so in a way we can measure both
quantitatively and qualitatively. As we discuss below, of the fifteen trading days following the
accident only three or four were days on which there was significant endorsement-related news; the
other days were largely quiet. Our Google search intensity data, which we describe below, confirm
this view by identifying clear peak periods of interest coinciding with the timing of endorsementrelated news. The endorsement-related activity lags the onset of the broader scandal significantly;
for example, Google searches for “Tiger Woods endorsement” did not take off until a few days
after the accident, did not peak until ten trading days after the accident, and experienced a third
bump on December 14, 2009. Qualitatively identifiable “news days” along with (similarly timed)
quantitative measurements of endorsement-related interest allows us to ask whether the pattern of
stock price changes during the scandal matches the pattern of news/interest.
The large number of sponsors also allows us to augment the analysis by collecting data for a wide
set of competitors to Tiger Woods’ sponsors. These data are useful in several ways. They allow us
to control for industry-specific factors affecting sponsors’ stock prices, because to the extent that
competitors and sponsors share industries those factors should also change stock prices for com5
In this respect, our work follows that of Farrell, Karels, and Montfort (2000) and Mathur, Mathur, and Rangan
(1997).
6
petitors. More important, our competitor stock price data allow us to estimate whether sponsors’
losses after the scandal are competitors’ gains. Whether that is true depends on substitutability
between sponsors’ products and competitors’ products, and the extent to which celebrity endorsements create new demand, or merely steal business from competitors. Understanding whether
celebrity endorsement is business-stealing or pure value creation is important both conceptually
and for business strategy, but there has been very little empirical work examining the question.6 .
Finally, the dramatic nature of this particular scandal—an extremely damaging set of events for
the world’s leading endorser—allows us to examine the general role of reputation risk in determining
firm value for endorsement-intensive firms in general. As we discussed above, reputation risk
should be priced by the stock market. Following the Tiger Woods scandal, the media devoted
substantial attention to that risk; for example, a Google search for “celebrity reputation risk”
yields stories largely written about Tiger Woods after the scandal. One can argue that the scandal
either directly altered perceptions of the level of risk, or that it simply alerted the market to
precisely how important reputation risk can be for endorsement-intensive firms. In either event,
one might expect a stock market reaction. There is also evidence of a market response, by insurance
companies offering protection against celebrity reputation risk; a New York Times article written
January 31, 2010 was titled “Insuring Endorsements Against Athletes Scandals,” and stated this:7
In the wake of the Tiger Woods scandal, insurers are being inundated with inquiries
from corporations seeking to protect their investments, their brands and even their
sales when their celebrity endorsers suffer public embarrassment...In a new wrinkle,
more companies are trying to insure against the potential loss of sales when an athlete
product endorser is involved in a scandal.
Whether the scandal in fact changed market-level perceptions of reputation risk is of course an
empirical question. We explore that question by estimating post-scandal stock price changes for
two subsets of sponsors’ competitors: those who are themselves endorsement-intensive, and those
who are not endorsement-intensive. If the scandal sent a market-wide signal about reputation risk,
one might expect that risk to affect stock prices for all endorsement-intensive firms, even those
who do not have Tiger Woods as an endorser. We test that by comparing competitors’ returns
6
As we noted above, the exceptions are the work by Mathur, Mathur, and Rangan (1997), who find that competitors to Michael Jordan’s sponsors experience “very weak” stock price changes after Jordan’s return to professional
basketball, and the work by Chung, Derdenger, and Srinivasan (2011) showing that competitors of Nike gained golf
ball sales after the scandal
7
http://www.nytimes.com/2010/02/01/sports/01insurance.html.
7
for the two subsets: if market-wide perceptions about reputation risk changed, one would expect
that competitors with endorsement deals would fare relatively worse than competitors without
endorsement deals.
3
The Endorsement Deals of Tiger Woods and the Scandal
Prior to November 2009, Tiger Woods’ annual endorsement income was estimated to be roughly
$100 million, a figure roughly twice as large as that for any other athlete.8 We are able to identify
seven publicly owned, domestically traded companies with which Tiger Woods had an endorsement
or sponsorship deal as of November 27, 2009. We list those companies in Table 1.9 While the
details of most contracts are private, the five most valuable contracts were seemingly with Accenture, Gillette, Nike, PepsiCo (Gatorade) and Electronic Arts (EA).10 Those five deals generated
approximately $80-90 million in annual income prior to the scandal. In the empirical work below,
we estimate some stock price effects for this subset of “Primary” firms.
It is also worth noting that some sponsors augment the endorsement relationship by making
complementary co-investments in product lines, brand name or other assets, the value of which
might also be tied to the endorser’s reputation. There are three such firms in our sample. Nike
has a considerable complementary investment in the Nike golf product line, which did not exist
prior to the Tiger Woods endorsement contract. Electronic Arts sells the “EA Tiger Woods” line
of video games, and recently launched a new “Tiger Woods Online” video game. Gatorade invested
considerable resources in developing a “Tiger Focus” drink.
We draw this distinction because for firms with such co-investments linked to the “Tiger brand,”
the link between reputation risk and firm value could go beyond the dollar value of the endorsement
contract and its short-run effect on sales/profits. Developing and marketing a new product line
requires a considerable up-front investment, as well as substantial production and marketing costs.
The Nike golf line, for example, is a brand with considerable asset value, accumulated via Nike’s
substantial up-front and ongoing investment in R&D, physical capital and brand equity. So, for
firms with such complementary investments, changes in stock prices will reflect changes in the
value of those assets, as well as changes in direct sales associated with the endorsement deal. In the
8
http://sportsillustrated.cnn.com/more/specials/fortunate50/2009/.
See http://web.tigerwoods.com/sponsors/sponsors for a complete list. Some of the companies on that list are
either privately held, or traded on foreign exchanges; we do not track those companies.
10
See http://industry.bnet.com/advertising/10005016/the-tiger-woods-sponsor-deathwatch-at-nike-digs-in-heels/
for details.
9
8
empirical work below we estimate stock price effects for the “Tiger Brand” group of Nike, Electronic
Arts and Gatorade: the set of firms with substantial complementary investments associated with
Tiger Woods.
3.1
The Timeline of the Scandal
The scandal began with a car accident on the evening of November 27, 2009—a Friday, meaning
that the first trading day after the release of “news” was Monday November 30, 2009.11 Following
the night of the accident, several potentially reputation-damaging pieces of information emerged,
primarily involving extramarital affairs. Events culminated ten trading days later (December 11,
2009) with Tiger Woods’ announcement of an ‘indefinite leave’ from golf.12 Table 2 summarizes
these events day-by-day, starting one week before the scandal, and ending on December 18, 2009—
fifteen trading days after the accident. Beyond the fifteen trading-day horizon we lose statistical
precision, so we confine ourselves to this window rather than some longer time period.
As illustrative evidence regarding the rise and decline of media interest in the story, we examine
the results of Google Insights searches related to the scandal.13 Previous work [e.g., Da, Engelberg,
and Gao (2011) and follow-on studies] has shown that Google search intensity is correlated with
stock price changes, implying that search intensity captures investor attention. Google’s Insights
data quantify internet interest in a subject on a 100-point scale, as measured by the popularity
of keyword searches. Data are normalized search-by-search, with 100 representing peak activity
during the search period. To be clear, the scale is informative within a search rather than across
searches: within a particular search “100” always implies twice as much search activity as “50,”
but the peak values of 100 across two different searches may represent different absolute levels of
interest.
The most popular three-word search terms following the scandal were “Tiger Woods accident”
and “Tiger Woods wife.”14 Figure 1 shows interest in these terms starting on November 26,
2009 and ending on December 18, 2009. Interest in these topics was at zero according to Google
Insights before then, suggesting that the pre-accident National Enquirer allegation was not taken
11
For a timeline and some details about the allegations, see http://www.montrealgazette.com/sports/timeline+
Tiger+Woods+decline/3374668/story.html.
12
One piece of scandal-related news predates the accident by four days: allegations of an affair in the National
Enquirer, released on November 23, 2009. We consider the possible effect of that early news in the empirical work
below, and find that it does not appear relevant.
13
One can find the search page here: http://www.google.com/insights/search/.
14
We observe this by starting with a general search for “Tiger Woods.” Given a general starting search, Google
Insights shows a rank ordering of the most popular refined search terms associated with the general search.
9
particularly seriously. Interest in the “accident” search peaks on the day of the accident, then dies
out quickly. Interest in the “wife” search builds after the accident and peaks on December 2-3,
the latter date being that on which Tiger Woods issued a statement admitting “transgressions.”
Interest in the “wife” search diminishes until a resurgence on December 8th, then falls again. By
December 18th, interest appears to have fully waned. Data over longer post-scandal windows show
no resurgence in interest over the next two years.
3.2
The Scandal and Sponsor Firms
Returning to Table 2, we also document endorsement-related news during the scandal. Endorsementrelated announcements lag general news about the scandal; the first piece of endorsement-related
news came on December 3, when Nike and Gillette issued press releases confirming support for
Woods. On December 8, Gatorade announced cancellation of its Tiger Woods-branded sports
drink; the announcement came late in the day, after the close of trading.15 The next pieces of news,
clustered on December 11 and over the following weekend, include Accenture dropping Woods, and
Gillette announcing that it would “limit” Woods’ role in marketing going forward. These pieces of
information coincide with the announcement on December 11 of Woods’ leave from golf. While we
do not extend the window of our analysis beyond December 18 because we have limited statistical
power after then, it is perhaps worth noting that AT&T dropped Tiger Woods on December 31,
2009, and Gatorade dropped Woods on February 26, 2010.
Figure 1 sheds light on the relative importance of these events by plotting Google search intensity
for the term “Tiger Woods endorsement.” That search term takes a value of zero until the day after
the accident, and has its first spike on December 3—the Nike/Gillette press release day. Its peak
is on December 8/9 following the Gatorade announcement, and interest remains high until after
the announcement on December 13 that Accenture was dropping Woods. While the correlation is
not perfect, it is high—Google intensity corresponds closely to the pattern of endorsement-related
announcements following the scandal.
As further suggestive evidence that the scandal mattered for sponsor firms, we show in Figure 2
the average Google search intensity for our seven sponsor firms between January 2009 and January
2010. We construct two averages. One average uses search intensity based on parent company name
and the other uses search intensity based on the brand name endorsed by Tiger Woods (see Table
15
Whether cancellation was in the offing prior to the scandal is an open question.
http://www.cnbc.com/id/34330134/Gatorade Tiger Discontinued Not Tied To Events Company
http://www.dailyfinance.com/2009/12/09/gatorade-drops-tiger-woods-endorsed-drink/.
10
See,
and
1 for details). This distinction matters only for two of the seven sponsors (Pepsi/Gatorade and
Proctor and Gamble/Gillette). Also, we use “Electronic Arts” as the search term for both parent
and brand, because a search for the Tiger Woods-themed golf video game (“Tiger Woods PGA Tour
Golf”) would spuriously capture broader searches for Tiger Woods. We weight all parents/brands
equally in the index.
The shaded area on the figure covers the two weeks of peak interest in the scandal. The brandspecific average peaks during that week, meaning that for our seven brands, this time period was
on average the period of greatest worldwide Google search interest over the preceding year. Three
of the seven brands in our sponsor group experience the peak (=100) of their 2009 search intensity
during the two weeks of the scandal, and AT&T peaks during the week of December 31, when it
announced dropping Woods.
The parent-specific pattern is similar, although there are three other time periods in which
parent-level intensity exceeds that during the scandal. The first comes during February 22-28, and
is driven by a 100 search intensity level for Accenture. That week coincides with the Accenture
Match Play Championship, a golf tournament in which Tiger Woods played, and a key part of
Accenture’s Tiger Woods-related marketing activities. A second peak comes during November
8-14, and is driven by a 100 intensity value for Electronic Arts, which announced a substantial
negative earnings report and layoffs during that week.16 The third peak is during September 20-26
and driven by Gillette; while we can find no corporate announcements by Gillette during that week,
the rock band U2 played a concert at Gillette Stadium in Foxborough, Massachusetts, which may
have driven spurious interest in “Gillette” as a search term.
Looking at the gap between the parent-specific and brand-specific average lines is also informative. The averages move together quite closely for nearly all of 2009, but deviate by the greatest
amount precisely at the peak of the scandal—when interest in the brands relative to the parents
would have been highest, based on affiliation with Tiger Woods.
All of this evidence points to a substantive qualitative relationship between the events of the
scandal, attention to endorsement values, and interest in sponsor firms. Google intensity correlates
quite closely with endorsement-related news, Google intensity for our sponsor firms correlates quite
closely with endorsement-related news, and prior work shows that search intensity is correlated
with changes in firm value. Our empirical work examines these links more formally.
16
The “mini-peak” in February 1-7 is also EA-driven and coincides with another negative announcement.
11
4
Estimated Stock Market Effects of the Scandal
To estimate stock price changes our set of sponsor firms and competitors following November 27
2009, we estimate an event study. Our method is standard in marketing, economics and finance,
and as we discuss above has been employed previously in studies linking stock market value to
celebrity endorsements.17
Our primary specification is:
Rit = αi + βim Rtm +
X
δs Dst + it ,
(1)
s
where,
• Rit = the return on shares of sponsor i at time t,
• Rtm = the return on the Dow Jones value-weighted total market index at time t,
• δs = the abnormal return on day s after the accident,
• Dst = a dummy variable equal to one during day s after the accident,
• it = an error term.
The specification is a standard market model where the dependent variable is a sponsor’s daily
percentage return exclusive of dividends, from Wharton Research Data Services and the Center
for Research in Stock Prices (CRSP). The independent variables include a value-weighted total
market return. The model allows for sponsor-specific daily mean returns (alphas) and correlations
with market/competitor returns (betas). Our estimation window begins three months before the
accident date and extends to December 18, 2009. Event date “zero” is November 27, and November
30, 2009 is the first trading day after the event date.
Our model yields estimates of daily abnormal returns, δs , which are deviations of actual returns
on the days after the scandal from those predicted by the model. We weight observations by market
capitalization, effectively estimating the abnormal returns that one would earn by holding a valueweighted portfolio of Tiger Woods’ sponsors.18 We also estimate cumulative abnormal returns
(CARs)—which are running sums of the daily abnormal returns—starting on November 30th. The
CARs estimate sponsors’ total loss over a multi-day window starting on event date one, relative
17
See, e.g., MacKinlay (1997) for a survey.
Estimating a value-weighted return is more informative than estimating an equally-weighted return, because
total dollar gains or losses for shareholders depend on the value-weighted average return. We use daily market
capitalization to construct the weights. Results are identical if we use weights as of the event date, or averaged over
the month prior to the event date.
18
12
to the market returns. In the results below we report abnormal returns and CARs for windows
extending up to fifteen trading days after the event date.
When examining the effect of a single event on multiple firms, it is important to adjust the
estimated standard errors for the contemporaneous correlation of sponsor-specific errors on the
same day. We use the procedure in Salinger (1992) for calculating standard errors on the cumulative
abnormal returns. The procedure involves making a simple transformation to the data matrix that
yields correct standard errors. We also omit observations for the week preceding November 30,
2009. Including them does not change the results, and we find no evidence of pre-event abnormal
returns.
We also estimate specifications including a value-weighted portfolio of competitors’ returns:
c
Rit = αi + βim Rtm + βic Rit
+
X
δs Dst + it ,
(2)
s
where all variables are defined as above and
c = the return on shares of sponsor i’s competitors at time t,
• Rit
The competitor portfolio includes the first ten firms listed by Google Finance as “competitors”
of the sponsor—meaning the sponsor’s parent company.19 Table A.1 lists competitors for each
sponsor; we include only competitors traded on U.S. stock exchanges.
From a methodological perspective, whether to include competitors’ returns is ambiguous. The
advantage of including competitors’ returns is that it can control more completely for confounding
industry-specific contemporaneous influences on sponsors’ stock prices. With this in mind we
classify competitors based on the parent company rather than the sponsor brand; the ideal match
for a particular parent company is another firm that competes in a set of industries identical to
those of the sponsor’s parent company. Of course, in some cases (e.g., Accenture) the sponsor
brand and parent company are identical.
Including competitors’ returns may be less correct, on the other hand, if the scandal itself
affected competitors’ returns. In that case, our estimated abnormal returns for sponsors would
be measured relative to competitors rather than the entire market. Conceptually, this changes the
interpretation of the estimated abnormal returns from answering “How did sponsors fare relative to
the market?” to answering “How did sponsors fare relative to the market and their competitors?”
19
We have estimated the model using the first five or three competitors, and also using the Yahoo! Finance
competitor list. Varying the specification of competitors’ returns has no effect on the results. Nor does weighting
competitors’ returns equally.
13
The first question implies a counterfactual in which an investor holds the market portfolio, while the
second implies a counterfactual portfolio with greater weight on sponsors’ industries. The second
approach also would complicate a comparison of sponsors’ to competitors’ abnormal returns during
the same period, as it would double-count (or zero out) the effect of changes in competitors’ values.
For the latter reason in particular, in most of the empirical work below we will present results that
estimate sponsors’ abnormal returns relative to the market (i.e., the abnormal returns estimated
from equation 1 above). We do show the results relative to competitors’ returns for completeness.
In some cases we are interested in examining abnormal returns that vary across firms within
the same day. We estimate those using the more flexible specification:
Rit = αi + βim Rtm +
X
δis Dst + it ,
(3)
is
This more flexible specification allows us to conduct non-parametric sign and rank tests regarding the post-event abnormal returns δis . In both tests the null hypothesis is that post-event
abnormal returns are centered on zero, which is what one would expect if the post-event period
contains no systematic news about firm value. Rejecting the null suggests that some (either positive or negative) information affected sponsor firms’ returns. In these models we also correct for
contemporaneous correlation of errors across sponsor firms.
4.1
Primary Results
Table 3 shows estimates of cumulative abnormal returns (CARs) for all sponsors, for the Primary
group only (Nike, Gatorade, Electronic Arts, Accenture and Gillette) and for the Tiger Brand
group (Nike, Gatorade and Electronic Arts). The first three columns show full results for the
model in equation (1) including the market return, and the second three columns show results
from the model controling for both market and competitor returns in equation (2). The fit in
the second three columns is better, and the competitors’ returns explain a share of variation in
sponsors’ returns that is both statistically and economically significant.20 The fact that sponsors’
and competitors’ returns co-move before the event, and materially so, suggests that our definition
of “competitor” captures firm- or industry-specific similarity across firms that the market return
does not capture—i.e., that competitors are usefully defined.
In every model the point estimates are fairly flat and not statistically significant until eight
20
The p-value on an exclusion restriction for competitors’ returns is less than 0.0001.
14
trading days after the accident; in every instance the estimates CARs turn sharply negative and
remain so until the end of our fifteen-day event window. By and large the estimates are statistically
significant, particularly later in the event window and for the Primary and Tiger Brand sub-samples.
The point estimates for the smaller sub-groups are also larger (more negative). Referring to the
first three columns, in the Primary subsample the 10-(15-)day CAR shows a loss of 3.0%(5.3%),
and in the Tiger Brand subsample the 10-(15-)day CAR shows a loss of 3.4%(5.8%).
The results in the second three columns, from the model including competitors’ returns, show
a similar pattern. The similarity of the point estimates suggests that average competitor returns
did not move much following the scandal, although below we decompose those average returns and
show that certain competitors benefited while others lost value during the scandal.
Figure 3 and Figure 4 provide graphical detail on the pattern of losses over time. Figure 3 shows
the cumulative abnormal returns for our three sponsor groups, while Figure 4 shows individual
CARs for the Primary group. We do not show CARs for the full set of firms because the CAR
for TLC Vision is extremely large and negative, reducing the viewing scale of CARs for all other
firms. The large negative CAR for TLC almost certainly foreshadows its bankruptcy declaration
on December 21 2009. For our weighted average CARs in the full sample this does not matter much
because TLC’s weight in the portfolio is trivially small, but it is worth noting. If one weights the
portfolio equally, the CARs for portfolios including TLC become more negative after the scandal.
The graphical representations of CARs illustrate two points. One is that CARs turn negative
only near the date of the first negative endorsement-related announcement on December 8 2009
(the Gatorade drink cancellation). This is suggestive evidence that our results capture the effects
of endorsement-related news. A second point, from Figure 4, is that even Accenture, which has
the most positive CAR during the scandal, experiences a decline in value after the first negative
endorsement-related news is released.
Table 4 shows daily abnormal returns and presents the results of the sign and rank tests. The
main body of the table shows daily abnormal returns for each of our main sponsor groups. These
abnormal returns are the individual δs coefficients, which are averaged across firms (weighted by
firm value). One can see that the largest negative returns occur in two clusters, 3-4 and 8-9 trading
days after the onset of the scandal, corresponding to December 2-3 and December 9-10 respectively.
The bottom four rows use the firm-specific daily abnormal returns δis (not shown in the table)
to conduct both sign and rank tests over 10-day and 15-day windows. Again, the null hypothesis
in these tests is that returns are centered on zero. The alternative (one-tailed) hypothesis in
15
each test is that the returns are centered on a negative value, indicating the systematic release
of negative information affecting all firms. The sign test uses only information about the sign
(positive or negative) of each coefficient, while the rank test uses information about both signs and
magnitudes. For the full sponsor group, the p-values for both sign tests are below 0.10. Results
for the subsamples are more significant. For the Primary and Tiger Brand both sign test p-values
are below 0.05. The pattern for the rank tests is similar. In all, these results provide strong
evidence that abnormal returns after the scandal are systematically negative, particularly for the
Primary/Three groups.
4.2
Endorsement-related News, Search Intensity and Abnormal Returns
We now tie the day-by-day pattern of abnormal returns from Table 4 to the patterns of news/search
behavior we documented in Table 2 and Figure 1. This analysis corroborates the view that our
estimated abnormal returns are related to the Tiger Woods scandal, rather than some other factor(s).
We first approach the question graphically. Figure 2 plots the Google Insights index for “Tiger
Woods endorsement” over trading days 1-15 in the event window. On the same axis we also
plot the negative of average abnormal returns for sponsor firms over the event window, using our
standard groupings of sponsors; each point on the figure corresponds to one coefficient from the
first three columns of Table 4. Plotting the negative of abnormal returns makes easier the visual
comparison between higher (more positive) search intensity and larger (more negative) abnormal
returns for sponsor firms.21 The figure shows a strong link between Google search intensity and
daily abnormal returns. Search intensity peaks on December 9, and that is the day with the largest
(negative) abnormal return for any group of firms. Day-by-day movements up/down in search
intensity also correlate with abnormal returns.
We next undertake a more formal statistical analysis linking endorsement-related news/search
intensity to the magnitude of abnormal returns. The model for this analysis is:
δˆis = α + βN ewss + is ,
(4)
where,
21
Specifying the relationship this way maintains the assumption that all news during the event window had a
negative effect. We have also constructed, but do not present here, a figure correlating search intensity with the
absolute value of returns; that assumes that search intensity could lead to large abnormal returns in either direction.
That figure looks quite similar, because most of the largest daily returns are negative.
16
• δˆis = the estimated abnormal return on shares of firm i on event date s from equation (1),
• N ewss = a time-varying measure of news/search intensity,
• is = an error term.
With seven firms and fifteen trading days during the event window, we have a total of 105
observations for these regressions when all sponsor firms are included, and 75/45 observations for
the Primary/Tiger Brand subsamples.
To fully explore the relationship between search/news intensity and abnormal returns, we use
three different measures of news/search intensity. The first is the level of search intensity for “Tiger
Woods endorsement” from Google Insights, as shown on Figures 1 and 5, re-scaled to be between
zero and one (rather than between 1 and 100). This takes on a minimum value of 0.07 (on November
30) and a maximum value of 1.00 (on December 9). Our second measure of search intensity is an
indicator taking on a value of one on days with a Google Insights score above 25 (0.25), and zero
otherwise; that occurs on December 2, 3, 8, 9, 10, 11, 14, 15 and 17 of 2009.22 Finally, we include a
qualitative indicator, self-defined, equal to one on the “endorsement-related news days” identified
in Table 2: December 3, 8, 9 and 14.
Table 5 presents results from these models. With every specification of news/search intensity,
the coefficients show more negative abnormal returns on days of greater news/search intensity. The
effects are larger for the Tiger Brand firms than for the sample as a whole. In the first set of rows,
the point estimates imply negative abnormal returns of 0.7%-2.6% on days with search intensity
equal to 1.00, relative to days with search intensity equal to 0.00. The second set of rows shows
negative abnormal returns of 0.4%-1.0% on days with search intensity greater than 0.25. And
finally, the coefficients in the last set of rows imply negative abnormal returns of 0.2%-1.4%.
Two important patterns emerge in these results. First, the correlation between news/search
intensity is much stronger for the “Tiger Brand” firms than for the other firms in the set of
sponsors—note the significantly higher r-squared terms in the last column of results. This is what
one would expect if the results reflect the downside of the scandal and the Tiger Brand firms had
more at stake. Second, and perhaps more important, our objectively measured search intensity
variable (the Google Insights measure) significantly outperforms our qualitative and subjectivelydefined “news day” measure, in terms of fitting the pattern of abnormal returns. This is a promising
22
We have also tried other cutoffs such as 0.50 or 0.75, or sets of indicators based on quartile cutoffs; the results
are qualitatively similar. We present results using the 0.25 cutoff here because it is generous relative to the other two
measures in the table in terms of classifying “high” intensity, and therefore provides a useful comparison to those
narrower measures.
17
result in the context of event studies that attempt to explain abnormal returns, because the Google
Insights-based variable avoids issues related to researcher-defined measures of which days after an
event are “important.”
As a robustness check, we show in Table A.2 the results of similar models that use other search
terms related to the scandal. The first two sets of rows use the search intensity for the “Tiger Woods
accident” search, and the second sets of rows use the “Tiger Woods wife” search.23 We show two
specifications for each alternative search measure: one including the “Tiger Woods endorsement”
search intensity (from Table 4) and one omitting that variable.
The results show quite clearly that while endorsement-related search intensity correlates quite
strongly with sponsors’ abnormal returns, non-endorsement-related but still scandal-related search
intensity is unrelated to the pattern of abnormal returns. The more general scandal-related search
terms are closer to zero in point terms, and never statistically significant. Furthermore, their
inclusion leaves the magnitude and significance of the endorsement-related coefficient unchanged.
This provides further evidence that our findings reflect something specific to the endorsementrelated effect of the scandal.
4.3
Competitor Returns and Endorsement Intensity
In this section, we examine returns for our sponsors’ competitors. For each of the seven firms in
our sponsor sample we collect daily return data for ten competitors, meaning that we examine
returns for as many as seventy competitors in the work below. Some competitors move in or out
of the sample during the estimation window, are not traded on a U.S. exchange, or are one of our
sponsors, meaning that we do not always have data for all seventy firms.
Our analysis of competitors’ returns focuses on two questions. First, we ask whether the scandal
appears to generate abnormal returns for competitors. One might imagine that losses for sponsor
firms could be gains for rivals of sponsors, if celebrity endorsements lead to business-stealing and
that business-stealing reverses after a scandal. Alternatively, it is possible that losses for sponsors
would not affect competitors’ returns, if celebrity endorsements simply create new value in a market
(perhaps relative to other markets, perhaps not). It might even be possible that one firm’s losses
could spill over to all competitors in a “category,” although this is perhaps more plausible for some
products (e.g., golf balls) than for others (e.g., sports drinks).
23
We have also estimated the model using the more general “Tiger Woods” search; that tracks “Tiger Woods wife”
quite closely and yields similar results.
18
A second question is whether those competitors who are themselves endorsement-intensive,
meaning that they also use celebrity endorsements as part of their marketing efforts, fared differently
from those competitors with no links to celebrities. The purpose of the second test, as we note
above, is to test for broader impacts of the Tiger Woods scandal. Given the prominence of Tiger
Woods as an endorser and his arguably impeccable reputation prior to the scandal, it is possible
that the scandal sent a negative market-wide signal about risk associated with any endorsement
deal. We classify a competitor as “endorsement intensive” if a Google search for “[competitor
name] celebrity endorsement” reveals that the competitor has at least one celebrity endorsement
deal during our event window. Table A.1 lists our competitors and whether we classify them as
endorsement-intensive. This is probably conservative, in the sense that relatively few of these firms
are as endorsement-intensive as the large firms that Tiger Woods endorses.24
The model for this analysis is the standard market model:
c
Rit
= αi + βic Rtm +
X
δsc Dst + it ,
(5)
s
where,
c = the return on shares of competitor i at time t,
• Rit
• Rtm = the return on the CRSP equally-weighted portfolio at time t,
• δsc = the abnormal return for competitor i from day s after the accident,
• Dst = a dummy variable equal to one during day s after the accident,
• it = an error term.
The specification allows for competitor-specific daily mean returns (alphas) and correlations
with market returns (betas). We weight the returns by competitor value (market capitalization).
We estimate competitors’ returns for all competitors, as well as competitors to the Primary/Tiger
Brand groups.
Table 6 shows ten-day CARs for the competitor sample. The first three columns show returns
for competitors who are not endorsement-intensive. Competitors’ CARs are positive, and rise as
sponsors’ returns fall—with the greatest changes occurring by day ten. This pattern dovetails with
the gradual onset of negative CARs for sponsors, and with the timing of endorsement-related news,
corroborating the view that competitors’ gains are sponsors’ losses. The point estimates grow in
size as we restrict the sample to competitors of the Primary and Tiger Brand groups, which is also
24
We experimented with several ways of classifying endorsement intensity, with little variation in the qualitative
results.
19
consistent with the pattern of sponsors’ losses.25
The more interesting results are those in the next six columns, which show that endorsementintensive competitors fared significantly worse than non-intensive competitors. The middle three
columns show that after event day two non-intensive competitors’ returns turn negative, and are
statistically significantly negative on day 14. More important, the difference between returns for
endorsement-intensive and non-intensive competitors is economically meaningful and statistically
significant, at least for the Primary/Three sub-samples. For the Primary subsample the CARs
are significant at 10% or 5% on all days after trading day 8 (December 9 2009), and range from
-2.1% to -3.3% in point terms, meaning that endorsement-intensive competitors lost roughly 2-3%
of value relative to their non-intensive competitors. The point estimates are larger for Tiger Brand
competitors but less significant statistically, reflecting the smaller sample size.
The relative gains for competitors without endorsement deals suggest the losses for sponsor firms
were at least in part gains for competitors—in other words, that celebrity endorsements transfer
value across firms. But the fact that being endorsement-intensive was treated more harshly in the
market suggests a second effect—that the scandal sent a negative market-wide signal, as suggested
in the New York Times article above, about the possible downside of celebrity endorsements. For
endorsement-intensive competitors, the net effect of the business-stealing effect (a gain) and the
reputation risk effect (a loss) appears to be nearly a wash. If we pool all competitors the average
CARs for the pooled group are close to (and not significantly different from) zero.
To confirm that our findings for competitors are endorsement-related, we estimate a series of
regressions of the form
δˆcis = αc + β c N ewss ∗ Endorsei + is ,
(6)
where,
• δˆis = the estimated abnormal return on shares of firm i on event date s from equation (1),
• N ewss = a time-varying measure of news/search intensity,
• Endorsei = an indicator equal to one if the competitor is endorsement-intensive,
• is = an error term.
The specification mirrors that in Table 5; it estimates the link between abnormal returns and
25
The difference in sponsors’ abnormal returns when measured relative to competitors need not equal competitors’
abnormal returns. The net difference depends both on the level of competitors’ returns, and on the correlation
between sponsors’ and competitors’ returns.
20
measures of news/search intensity. We allow the relationships to differ for endorsement-intensive
and non-intensive competitors by including an interaction term.
Table 7 shows the results of these models. The broad pattern is of a positive and statistically
significant relationship between endorsement-related news/search intensity and abnormal returns
for the baseline set of non-intensive competitors, and a relationship for endorsement-intensive competitors that is significantly less positive and close to zero on net. The effects estimated in this table
are generally smaller than those estimated for sponsors. In short: competitors’ returns during the
scandal are greatest precisely when sponsors’ losses are greatest, unless the competitors themselves
are endorsement-intensive.
In the top rows, the All and Tiger Brand coefficients show a positive and significant relationship
between the continuously measured “Tiger Woods endorsement” search intensity variable and abnormal returns for non-intensive competitors. Those coefficients are also positive and significant for
all groups using the discrete “intensity>0.25” variable. They are smaller and less significant using
our qualitative self-defined “news days” variable. The pattern for the interaction terms is similar,
in terms of size and significance. The interaction terms measure the difference between returns for
non-intensive and endorsement-intensive competitors—the sum of the two coefficients measures the
net effect on endorsement-intensive competitors. We also observe, as we did with sponsors’ abnormal returns, that the quantitative intensity measures from Google Insights correlate more strongly
with abnormal returns than does our self-defined “endorsement-related news day” variable.
The overall pattern of results is summarized by Figure 4, which highlights the differences between our three groups of affected firms: sponsors, competitors with endorsement deals, and competitors without endorsement deals.26 The relative differences across the groups are economically
meaningful; the scandal appears to have had far-reaching and substantive effects on a large set of
firms.
While we do not report the results, we have also estimated a model that pools all sponsors
and competitors, and estimates overall (value-weighted) effects on the “category portfolio.” These
models show negative, small (less than 1%) and borderline statistically significant effects overall.
In other words, the net effect on this entire set of firms is a small and weakly significant loss in
value, with significant “within-category” transfers from sponsor firms to non-intensive competitors
of sponsor firms. These results are broadly consistent with the results in Chung, Derdenger, and
Srinivasan (2011) from the golf ball market.
26
The sponsor coefficients here are those from the model without competitors’ returns, to avoid double-counting.
21
4.4
Robustness Checks and Caveats
While we do not present them here, we have conducted a variety of robustness checks. We have
estimated the models using a variety of event windows, with little effect on the results.27 We have
estimated models that include the pre-event week, or drop up to a month’s worth of pre-event
data. We have varied the weighting scheme (using time-invariant market capitalization weights, for
example). These modifications to the specification do not change the results.
Another robustness issue arises because PepsiCo announced a negative earnings revision on
December 9, 2009, and one might worry that the announcement contaminates our results. In
unreported specifications (which we show in an earlier working paper version), we break our ‘Tiger
Brand’ subsample of EA, Nike into two groups: PepsiCo and the other two firms. The abnormal
return for PepsiCo on December 9 is indeed negative and significant, but so are abnormal returns
for the other two firms, and the point estimates are very close. While one cannot rule out a negative
stock price effect of the announcement for PepsiCo, the pattern of results is consistent with the
release on December 9 of bad news common to Nike, EA and PepsiCo. Furthermore, the p-values
for the sign and rank tests using only EA and Nike returns are both below 0.05 over the 10-day
window, and are much larger for PepsiCo, which experienced several fairly large positive returns
during the event window.
A final point concerns interpretation of the results. Ideally, one would want to interpret any
abnormal returns as measuring percentage changes in the expected value of future economic profits.
In our case that is hard, if not impossible, for a few reasons. Most of our sponsor firms are large
multi-product firms, for which Tiger Woods endorses only a single product; Nike produces many
products outside its golf line, for example. Nike’s stock price, of course, reflects expectations about
its profits from all business lines. So, the percentage change in profits will be weighted by the shares
of economic profits flowing from “Tiger-related” products and “non-Tiger-related” products. One
could proxy for those shares using dollar values of sales—Nike golf, for example, represents roughly
ten percent of annual sales for Nike—but there is no guarantee that shares of expected future profit
correspond to shares of current dollar sales. Another complicating factor is that if the scandal sent
a market-wide signal about celebrity reputation risk, then even the non-Tiger-related business lines
might suffer. That would be particularly true for a company like Nike, which is one of the most
celebrity endorsement-intensive firms in the world. For these reasons, we try to be as structure-free
27
An earlier working paper version and a later auxiliary table prepared for referees, available upon request, show
results.
22
in the econometric model as possible; the caveat is that our results should be taken as indicating
the direction and overall dollar value (percentage change × market capitalization) of abnormal
returns, but should not be taken as indicating percentage changes in Tiger-related economic profit.
This is particularly important when comparing gains/losses across firms, for which “Tiger-related”
gains/losses and shares of economic profits may be very different.
5
Discussion and Conclusion
The Tiger Woods scandal provides a unique opportunity to understand more about the relationship
between stock market value and celebrity endorsements. Our first result confirms a direct dimension
of that link: the market value of Tiger Woods’ sponsors fell substantively after the scandal broke,
relative to the market values of firms without such endorsement deals. That finding is informative
in the context of the mixed evidence from previous work.
Beyond that, we shed light on some previously under-studied aspects of the endorsement/stock
price relationship. Firms with substantial co-investments in new products linked to the “Tiger
brand” suffered greater declines in value, presumably reflecting declines in the asset values or
brand equity associated with those products. This result highlights a further downside risk of
pairing celebrity endorsements with endorser-specific investments in products or branding. We do
not estimate whether our results reflect long-run declines in value, due to the limited statistical
power of longer-run tests, but we have no evidence over as long as one month after the scandal
of any reversion in prices. The efficient markets hypothesis would suggest that markets should
immediately price the downside of scandals correctly on average; of course, that need not have
been the case in this specific instance. Further work using more data from a broader set of scandals
might be able to shed light on whether there is any systematic under-reaction or over-reaction to
celebrity scandals.28
We also relate novel auxiliary data from Google Insights to abnormal returns during the scandal.
The level of interest in the search term “Tiger Woods endorsement” explains nearly forty percent
of the variation across firms and days in abnormal returns following the scandal, and does so in an
intuitive way. The search intensity variable significantly outperforms our own qualitative measure
of which days were endorsement-newsworthy, suggesting promising avenues for future research.
28
We know of no work on that particular question, although previous work (see, e.g., Bernard and Thomas (1989)
and follow-on work) has shown that markets might under-react to other value-changing events such as earnings
announcements.
23
Our estimates of competitors’ gains represent new evidence regarding how far-reaching the stock
market effects of celebrity endorsements can be. Competitors to sponsor firms measurably gained
value after the scandal, relative to the rest of the market. That finding has implications for business
strategy, in that competitors’ endorsement deals are one more factor affecting firm value, and can
transfer value across firms.
Finally, we find compelling evidence that how competitors fared during the scandal depended on
whether they also had celebrity endorsers or not; this result is confirmed by the post-event relationship between competitors’ abnormal returns and endorsement-related news/search intensity. Along
with the anecdotal evidence regarding how the scandal altered perceptions of celebrity endorsement
reputation risk, this evidence suggests a regime change in how equity markets priced reputation
risk. Whether that regime change persists is an open question, but if insurance companies indeed
start offering “reputation risk insurance” then that view will have passed a convincing market test.
24
References
Agrawal, J. and W. Kamakura. 1995. “The economic worth of celebrity endorsers: An event study
analysis.” The Journal of Marketing 59:56–62.
Bernard, V. L. and J. K. Thomas. 1989. “Post-Earnings-Announcement Drift: Delayed Price
Response or Risk Premium?” Journal of Accounting Research 27 (1):1–36.
Chung, Kevin, Timothy Derdenger, and Kannan Srinivasan. 2011.
“The Economic Value of
Celebrity Endorsements: Tiger Woods’ Impact on Sales of Nike Golf Balls.” Working paper,
Carnegie Mellon University.
Clark, J., T. Cornwell, and S. Pruitt. 2009. “The impact of title event sponsorship announcements
on shareholder wealth.” Marketing Letters 20:169–182.
Cornwell, T. B., S. W. Pruitt, and R. V. Ness. 2001. “The value of winning in motorsports:
Sponsorship-linked marketing.” Journal of Advertising Research 41:17–31.
Da, Zhi, Joseph Engelberg, and Pengjie Gao. 2011. “In Search of Attention.” The Journal of
Finance 66 (5):1461–1499.
Ding, H., A. Molchanov, and P. Stork. 2008. “The value of celebrity endorsements: A stock market
perspective.” Marketing Letters :1–17.
Erdogan, Z. 1999. “Celebrity endorsement: a literature review.” Journal of Marketing Management
15:291–314.
Farrell, K. and W. Frame. 1997. “The value of olympic sponsorships: who is capturing the gold?”
Journal of Market-Focused Management 2:171–182.
Farrell, K., G. Karels, and K. Montfort. 2000. “Celebrity performance and endorsement value: the
case of Tiger Woods.” Managerial Finance 26:1–15.
Fizel, J., C. McNeil, and T. Smaby. 2008. “Athlete endorsement contracts: the impact of conventional stars.” International Advances in Economic Research 14:247–256.
Louie, Therese A., Robert L. Kulik, and Robert Jacobson. 2001. “When Bad Things Happen to
the Endorsers of Good Products.” Marketing Letters 12 (1):13–23.
25
MacKinlay, A. 1997. “Event studies in economics and finance.” Journal of Economic Literature
35:13–39.
Mathur, L., I. Mathur, and N. Rangan. 1997. “The wealth effects associated with a celebrity
endorser: The Michael Jordan phenomenon.” Journal of Advertising Research 37:67–73.
Mishra, D., G. B. Jr, and H. Bhabra. 1997. “Assessing the economic worth of corporate event
sponsorships: a stock market perspective.” Journal of Market Focused Management 2:149–169.
Miyazaki, A. D. and A. G. Morgan. 2001. “Assessing market value of event sponsoring: Corporate
olympic sponsorships.” Journal of Advertising Research 41:9–15.
Pruitt, Stephen W., T. Bettina Cornwell, and John M. Clark. 2004.
“The NASCAR Phe-
nomenon: Auto Racing Sponsorships and Shareholder Wealth.” Journal of Advertising Research
44 (03):281–296.
Salinger, Michael. 1992. “Standard Errors in Event Studies.” Journal of Financial and Quantitative
Analysis 27 (01):39–53.
Samitasa, A. and D. Kenourgiosb. 2008. “Athens’ olympic games 2004 impact on sponsors’ stock
returns.” Applied Financial Economics 18:1569–1580.
Till, B. and T. Shimp. 1998. “Endorsers in advertising: The case of negative celebrity information.”
Journal of Advertising 27 (67-82).
26
6
Tables
Table 1: Sponsor Firms, Endorsement Contracts and Market Value
Sponsor
Nike
Gatorade
Accenture
Gillette
Tiger Woods PGA Tour Golf
AT&T
TLC Laser Eye Centers
Parent Company
Endorsement value (/yr.)
Nike
$20-30 million
Pepsico
$20 million
Accenture
$20 million
Procter and Gamble
$15 million
Electronic Arts
$8 million
AT&T
n/a
TLC
n/a
Market Cap
$32 Billion
$95 Billion
$26 Billion
$179 Billion
$5.76 Billion
$165 Billion
$4.04 Million
Notes: We include all sponsors for which we can obtain stock price data. Market cap values are
as of mid-December 2009. AT&T's relationship with Woods involves sponsoring a golf
tournament and charity events, in exchange for product placement (e.g., on Tiger Woods' golf
bag).
27
28
1
2
3
Jaimee Grubbs and Kalike Moquin named as mistresses. Woods
issues first public statement admitting "transgressions."
Transcript of 911 call by neighbor released.
Date of accident
Trading
Day
Scandal-related news
National Enquirer report: affair with Rachel Uchitel.
Endorsement-related news
Notes: Dates in italics are weekend days (non-trading days). AT&T dropped Woods on December 31, 2009. Gatorade dropped Woods on February 26, 2010.
December 3, 2009
4
Nike, Gillette issue press releases confirming support for Woods.
December 4, 2009
5
December 5, 2009
Jamie Jungers comes forward as mistress.
December 6, 2009
Cori Rist and Mindy Lawton named as mistresses.
December 7, 2009
6
Holly Sampson named as mistress, bringing total to seven.
December 8, 2009
7
Woods' mother-in-law rushed to the hospital from Woods' home.
Gatorade drops Tiger Woods-branded drink (after close of trading).
December 9, 2009
8
December 10, 2009
9
December 11, 2009
10
Woods announces he will take an "indefinite" leave from golf
Accenture removes Woods' image from its Web site.
December 12, 2009
Gillette announces that it is "limiting" Woods' role in marketing.
December 13, 2009
Accenture drops Woods.
December 14, 2009
11
December 15, 2009
12
December 16, 2009
13
December 17, 2009
14
December 18, 2009
15
Sources: http://sports.espn.go.com/golf/news/story?id=4922436 and http://gawker.com/5421795/the-tiger-woods-saga-a-definitive-timeline.
http://abcnews.go.com/GMA/experts-tiger-woods-valuable-drop-sponsors/story?id=9236260
http://socyberty.com/people/nike-ea-sports-pg-to-drop-tiger-woods-endorsements/
Date
November 23, 2009
November 24, 2009
November 25, 2009
November 26, 2009
November 27, 2009
November 28, 2009
November 29, 2009
November 30, 2009
December 1, 2009
December 2, 2009
Table 2: Chronology of Scandal- and Endorsement-related News
Table 3: Cumulative Abnormal Returns for Sponsor Firms
Relative to:
Days after event All Firms
One
-0.004
(0.004)
Two
0.001
(0.005)
Three
0.003
(0.007)
Four
0.001
(0.008)
Five
0.003
(0.009)
Six
0.007
(0.010)
Seven
0.000
(0.010)
Eight
-0.008
(0.011)
Nine
-0.007
(0.012)
Ten
-0.009
(0.013)
Eleven
-0.010
(0.014)
Twelve
-0.019
(0.014)
Thirteen
-0.023
(0.015)
Fourteen -0.028*
(0.016)
Fifteen -0.032*
(0.016)
Observations
605
R-squared
0.291
Market only
Market, competitors
Primary Tiger Brand
-0.004
-0.004
(0.004)
(0.006)
0.001
0.012
(0.006)
(0.009)
0.001
0.007
(0.008)
(0.011)
-0.007
-0.004
(0.009)
(0.013)
-0.005
0.004
(0.010)
(0.014)
-0.006
0.008
(0.011)
(0.016)
-0.011
0.000
(0.012)
(0.017)
-0.022
-0.027
(0.013)
(0.018)
-0.024*
-0.027
(0.014)
(0.020)
-0.030**
-0.034
(0.015)
(0.021)
-0.031*
-0.040*
(0.016)
(0.022)
-0.039**
-0.038
(0.017)
(0.023)
-0.042**
-0.044*
(0.018)
(0.024)
-0.047**
-0.051**
(0.019)
(0.025)
-0.053***
-0.058**
(0.019)
(0.027)
435
261
0.314
0.336
All Firms
-0.002
(0.004)
0.001
(0.005)
0.003
(0.006)
0.001
(0.007)
0.005
(0.008)
0.010
(0.009)
0.002
(0.010)
-0.009
(0.011)
-0.010
(0.012)
-0.011
(0.012)
-0.011
(0.013)
-0.018
(0.014)
-0.020
(0.014)
-0.021
(0.015)
-0.024
(0.016)
605
0.360
Primary
-0.003
(0.004)
-0.001
(0.006)
-0.002
(0.008)
-0.009
(0.009)
-0.007
(0.010)
-0.007
(0.011)
-0.012
(0.012)
-0.022*
(0.013)
-0.027**
(0.014)
-0.035**
(0.015)
-0.034**
(0.015)
-0.042**
(0.016)
-0.042**
(0.017)
-0.042**
(0.018)
-0.047**
(0.019)
435
0.375
Tiger Brand
-0.002
(0.006)
0.009
(0.008)
0.006
(0.010)
0.001
(0.012)
0.008
(0.013)
0.011
(0.015)
0.003
(0.016)
-0.024
(0.017)
-0.029
(0.018)
-0.039**
(0.020)
-0.042**
(0.021)
-0.040*
(0.022)
-0.040*
(0.023)
-0.037
(0.024)
-0.041*
(0.025)
261
0.433
Notes: Coefficients are cumulative abnormal returns (CARs) weighted by firm value (market
capitalization). First three columns show results of the market model in equation (1). Second three
columns show results of the market model including competitors' returns. Event date is November
27, 2009. Estimation window begins three months before event date, and ends one week before
event date. Standard errors are adjusted for contemporaneous correlation across firms. "All firms"
include all listed in Table 1. "Primary" includes Nike, EA, Accenture, PepsiCo (Gatorade) and P&G
(Gillette). "Tiger Brand" includes Nike, EA and PepsiCo. Numbers in parentheses are standard
errors. Asterisks indicate significance at 10% (*), 5%(**) and 1%(***) or better.
29
Table 4: Average Daily Abnormal Returns for Sponsor Firms
Days after event
One
Observations
R-squared
10-day sign test p-value
All Firms
-0.004
(0.004)
0.005
(0.004)
0.002
(0.004)
-0.002
(0.004)
0.001
(0.004)
0.004
(0.004)
-0.007*
(0.004)
-0.008**
(0.004)
0.000
(0.004)
-0.001
(0.004)
-0.001
(0.004)
-0.009**
(0.004)
-0.003
(0.004)
-0.005
(0.004)
-0.004
(0.004)
553
0.290
0.061
Primary
-0.004
(0.005)
0.006
(0.005)
-0.001
(0.005)
-0.008*
(0.005)
0.001
(0.005)
-0.000
(0.005)
-0.006
(0.005)
-0.011**
(0.005)
-0.002
(0.005)
-0.006
(0.005)
-0.001
(0.005)
-0.007
(0.005)
-0.003
(0.005)
-0.005
(0.005)
-0.007
(0.005)
395
0.316
0.026
Tiger Brand
-0.004
(0.006)
0.016**
(0.006)
-0.005
(0.006)
-0.011*
(0.006)
0.008
(0.006)
0.004
(0.006)
-0.008
(0.006)
-0.027***
(0.006)
-0.000
(0.006)
-0.007
(0.006)
-0.006
(0.006)
0.002
(0.006)
-0.006
(0.006)
-0.007
(0.006)
-0.006
(0.006)
237
0.329
0.008
15-day sign test p-value
0.030
0.013
0.013
10-day rank test p-value
15-day rank test p-value
0.136
0.036
0.025
0.014
0.004
0.017
Two
Three
Four
Five
Six
Seven
Eight
Nine
Ten
Eleven
Twelve
Thirteen
Fourteen
Fifteen
Notes: Coefficients are abnormal returns weighted by firm value, estimated using the
model in equation (1). Event date is November 27, 2009. Standard errors are adjusted for
contemporaneous correlation across firms. "All firms" include all listed in Table 1.
"Primary" includes Nike, EA, Accenture, PepsiCo (Gatorade) and P&G (Gillette). "Tiger
Brand" includes Nike, EA and PepsiCo. Numbers in parentheses are standard errors.
Asterisks indicate significance at 10% (*), 5%(**) and 1%(***) or better. Shaded cells
indicate negative values. Sign and rank tests p-values use the full set of firm-day-specific
abnormal returns, estimated using the model in equation (2). For the sign and rank tests
the null hypothesis is that returns are centered on zero.
30
Table 5: Sponsors’ Abnormal Returns and News/Search Intensity
Dependent variable: Firm-level daily abnormal return
All Firms
Primary
Tiger Brand
"Tiger Woods endorsement" search intensity
-0.007*
(0.003)
-0.007*
(0.004)
-0.024***
(0.006)
Constant
0.001
(0.001)
0.048
-0.001
(0.002)
0.056
0.006*
(0.003)
0.306
"Tiger Woods endorsement" search intensity>25
-0.004*
(0.002)
-0.004*
(0.002)
-0.010**
(0.003)
Post-accident dummy
0.000
(0.001)
0.046
-0.001
(0.001)
0.057
0.002
(0.002)
0.182
Endorsement-related news day
-0.001
(0.002)
-0.004
(0.002)
-0.012***
(0.003)
Post-accident dummy
-0.002
(0.001)
0.006
105
-0.002*
(0.001)
0.042
75
-0.000
(0.002)
0.228
45
R-squared
R-squared
R-squared
Observations
Notes: Coefficients are from model (3) in text, modeling the relationship between firm-level
daily abnormal returns during the period [November 30, December 18] and three different
measures of endorsement-related news intensity. First set of rows use the level of "Tiger Woods
endorsement" search intensity on a [0, 1] scale to measure endorsement-related news. Second
set of rows use an indicator equal to one if "Tiger Woods endorsement" intensity exceeds 0.25,
and zero otherwise. Third set of rows use indicator variables equal to one on December 3,
December 9, December 11 and December 14; see Table 1 for details. Numbers in parentheses
are standard errors. Asterisks indicate significance at 10% (*), 5%(**) and 1%(***) or better.
31
Table 6: CARs for Competitors, by Endorsement Intensity
Not endorsement intensive
Days after event All Firms
One -0.003*
(0.002)
Two -0.001
(0.002)
Three -0.001
(0.003)
Four 0.004
(0.003)
Five -0.000
(0.004)
Six -0.002
(0.004)
Seven -0.001
(0.005)
Eight 0.008
(0.005)
Nine 0.012**
(0.005)
Ten 0.011**
(0.006)
Eleven 0.011*
(0.006)
Twelve
0.006
(0.006)
Thirteen 0.006
(0.007)
Fourteen 0.003
(0.007)
Fifteen 0.009
(0.007)
Observations
3355
R-squared 0.327
Primary Tiger Brand
-0.002 -0.009***
(0.002)
(0.003)
0.004*
-0.006
(0.003)
(0.005)
0.006*
-0.009
(0.003)
(0.006)
0.011***
-0.006
(0.004)
(0.007)
0.011***
0.001
(0.004)
(0.007)
0.007
0.005
(0.004)
(0.008)
0.004
0.002
(0.005)
(0.009)
0.008
0.006
(0.005)
(0.010)
0.017*** 0.022**
(0.006)
(0.010)
0.019*** 0.026**
(0.006)
(0.011)
0.021***
0.021*
(0.006)
(0.012)
0.019*** 0.027**
(0.007)
(0.012)
0.019*** 0.029**
(0.007)
(0.013)
0.016**
0.028**
(0.007)
(0.013)
0.024***
0.022
(0.008)
(0.014)
2328
1106
0.370
0.470
Endorsement intensive
All Firms
-0.000
(0.003)
0.008*
(0.004)
-0.002
(0.006)
-0.001
(0.006)
-0.001
(0.007)
-0.007
(0.008)
-0.011
(0.009)
-0.008
(0.009)
-0.007
(0.010)
-0.010
(0.011)
-0.009
(0.011)
-0.008
(0.012)
-0.009
(0.012)
-0.022*
(0.013)
-0.008
(0.013)
869
0.327
Primary Tiger Brand
-0.000
-0.000
(0.003)
(0.004)
0.008*
0.011**
(0.004)
(0.005)
-0.002
0.002
(0.006)
(0.007)
-0.001
0.000
(0.006)
(0.008)
-0.001
-0.001
(0.007)
(0.009)
-0.007
-0.005
(0.008)
(0.009)
-0.011
-0.009
(0.009)
(0.010)
-0.008
-0.009
(0.009)
(0.011)
-0.007
-0.006
(0.010)
(0.012)
-0.010
-0.007
(0.011)
(0.013)
-0.009
-0.007
(0.011)
(0.013)
-0.008
-0.009
(0.012)
(0.014)
-0.009
-0.013
(0.012)
(0.015)
-0.022*
-0.029*
(0.013)
(0.015)
-0.008
-0.017
(0.013)
(0.016)
869
632
0.327
0.271
Difference
All Firms
0.003
(0.003)
0.009**
(0.004)
-0.000
(0.005)
-0.004
(0.006)
0.001
(0.007)
-0.002
(0.008)
-0.007
(0.008)
-0.012
(0.009)
-0.015
(0.010)
-0.017*
(0.010)
-0.016
(0.011)
-0.010
(0.011)
-0.010
(0.012)
-0.019
(0.013)
-0.010
(0.013)
4224
0.330
Primary Tiger Brand
0.002
0.009
(0.003)
(0.006)
0.004
0.017**
(0.005)
(0.008)
-0.008
0.011
(0.006)
(0.010)
-0.011
0.007
(0.007)
(0.012)
-0.010
-0.002
(0.008)
(0.013)
-0.012
-0.008
(0.008)
(0.014)
-0.013
-0.009
(0.009)
(0.016)
-0.013
-0.013
(0.010)
(0.017)
-0.021*
-0.027
(0.011)
(0.018)
-0.026** -0.032*
(0.011)
(0.019)
-0.026**
-0.026
(0.012)
(0.020)
-0.023*
-0.034
(0.012)
(0.021)
-0.024*
-0.040*
(0.013)
(0.022)
-0.033** -0.054**
(0.014)
(0.023)
-0.026*
-0.036
(0.014)
(0.024)
3197
1738
0.34
0.33
Notes: Coefficients are CARs weighted by firm value. "Competitors" are the first ten firms listed by Google Finance for
each sponsor firm; see Table A1. "All firms," "Primary" and "Tiger Brand" include competitors of each group. "Endorsement
intensive" firms are those for which a Google search of "[company name] endorsement deals" yields one or more hits
describing an endorsement deal during the event window. "Difference" columns show differences between endorsementintensive and non-intensive competitors. Event date is November 27, 2009. Estimation window begins three months before
event date, and ends one week before event date. Standard errors are adjusted for contemporaneous correlation across firms.
Numbers in parentheses are standard errors. Asterisks indicate significance at 10% (*), 5%(**) and 1%(***) or better.
32
Table 7: Competitors’ Abnormal Returns and News/Search Intensity
Dependent variable: Competitors' daily abnormal return
All Firms
Primary
Tiger Brand
"Tiger Woods endorsement" search intensity
0.010***
(0.002)
0.006**
(0.002)
0.007**
(0.002)
"Tiger Woods endorsement" search intensity x
Endorsement-intensive competitor
-0.009***
(0.002)
-0.007***
(0.002)
-0.006**
(0.002)
0.001
(0.001)
0.049
0.002***
(0.001)
0.028
-0.001
(0.001)
0.027
"Tiger Woods endorsement" search intensity>25
0.006***
(0.001)
0.006***
(0.001)
0.005***
(0.001)
"Tiger Woods endorsement" search intensity>25 x
Endorsement-intensive competitor
-0.006***
(0.001)
-0.006***
(0.001)
-0.005***
(0.001)
0.001
(0.001)
0.068
0.002*
(0.001)
0.073
-0.002*
(0.001)
0.055
Endorsement-related news day
0.004***
(0.001)
0.003*
(0.001)
0.001
(0.002)
Endorsement-related news day x
Endorsement-intensive competitor
-0.005**
(0.002)
-0.004*
(0.002)
-0.004*
(0.002)
Constant
0.003***
(0.000)
0.019
775
0.003***
(0.000)
0.011
589
-0.000
(0.001)
0.015
347
Constant
R-squared
Constant
R-squared
R-squared
Observations
Notes: Coefficients are from model of the relationship between competitors' firm-level daily
abnormal returns during the period [November 30, December 18] and three different measures of
endorsement-related news intensity. Interaction terms test for differential responses across
endorsement-intensive and non-intensive competitors (see Table A1). First set of rows use the
level of "Tiger Woods endorsement" search intensity (on a [0, 1] scale) to measure endorsementrelated news. Second set of rows use an indicator equal to one if "Tiger Woods endorsement"
intensity exceeds 0.25, and zero otherwise. Third set of rows use indicator variables equal to one
on December 3, December 9, December 11 and December 14; see Table 1 for details. Numbers in
parentheses are standard errors. Asterisks indicate significance at 10% (*), 5%(**) and 1%(***).
33
Figures
Figure 1: Post-Accident Search Intensity Related to Tiger Woods
100
90
80
70
60
50
40
30
20
"Tiger Woods accident"
"Tiger Woods wife"
12/18/09
12/17/09
12/16/09
12/15/09
12/14/09
12/13/09
12/12/09
12/11/09
12/10/09
12/9/09
12/8/09
12/7/09
12/6/09
12/5/09
12/4/09
12/3/09
12/2/09
12/1/09
11/30/09
11/29/09
11/28/09
0
11/27/09
10
11/26/09
7
"Tiger Woods endorsement"
Notes: Search intensity is from http://www.google.com/insights/search/. Search intensity is normalized within each term, with peak volume at 100 and lower numbers
representing percentage of peak volume. “Tiger Woods accident” and “Tiger Woods
wife” are the top-ranked searches listed by Google insights following an initial search
for “Tiger Woods.”
34
Figure 2: Average Search Intensity for Sponsor Firms, Jan. 2009 to Jan. 2010
75.00
70.00
65.00
60.00
50.00
1/4/09
1/18/09
2/1/09
2/15/09
3/1/09
3/15/09
3/29/09
4/12/09
4/26/09
5/10/09
5/24/09
6/7/09
6/21/09
7/5/09
7/19/09
8/2/09
8/16/09
8/30/09
9/13/09
9/27/09
10/11/09
10/25/09
11/8/09
11/22/09
12/6/09
12/20/09
1/3/10
1/17/10
1/31/10
55.00
Average (Parents)
Average (Brands)
Notes: Search intensity is from http://www.google.com/insights/search/. Figure plots
unweighted averages of search intensity for the seven sponsor brand/parent firms listed
in Table 1.
35
Figure 3: Cumulative Abnormal Returns for Sponsor Firm Groups
0.02
0.01
0
-0.01
-0.02
-0.03
-0.04
-0.05
All Firms
Primary
Tiger Brand
Notes: Cumulative abnormal returns (CARs) are from Table 3.
36
12/18/09
12/17/09
12/16/09
12/15/09
12/14/09
12/11/09
12/10/09
12/9/09
12/8/09
12/7/09
12/4/09
12/3/09
12/2/09
12/1/09
-0.07
11/30/09
-0.06
Figure 4: Cumulative Abnormal Returns for Individual Sponsor Firms
0.08
0.06
0.04
0.02
0
-0.02
-0.04
Nike
EA
Pepsi
Gillette
ACN
Notes: Cumulative abnormal returns (CARs) are from sponsor-by-sponsor event studies
based on specification in equation 1.
37
12/18/09
12/17/09
12/16/09
12/15/09
12/14/09
12/11/09
12/10/09
12/9/09
12/8/09
12/7/09
12/4/09
12/3/09
12/2/09
12/1/09
-0.08
11/30/09
-0.06
-0.005
20
-0.01
10
-0.015
0
-0.02
"Tiger Woods Endorsement" search intensity
Abnormal return, all firms
Abnormal returns, Primary
Abnormal returns, Tiger Brand
Negative of daily abnormal return
30
12/18/09
0
12/17/09
40
12/16/09
0.005
12/15/09
50
12/14/09
0.01
12/11/09
60
12/10/09
0.015
12/9/09
70
12/8/09
0.02
12/7/09
80
12/4/09
0.025
12/3/09
90
12/2/09
0.03
12/1/09
100
11/30/09
Search intensity
Figure 5: “Tiger Woods endorsement” Search Intensity and Daily Abnormal Returns
Notes: Search intensity is from http://www.google.com/insights/search/, as in Table 2.
Abnormal returns are plotted as the negative of coefficients from Table 4.
38
Figure 6: Sponsors’ and Competitors’ Cumulative Abnormal Returns
0.03
0.02
0.01
0
-0.01
-0.02
-0.03
-0.04
Primary
Primary Comps, Sponsor-intensive
12/18/09
12/17/09
Primary Comps, Non-intensive
Notes: Coefficients are cumulative abnormal returns from Tables 3 and 6.
39
12/16/09
12/15/09
12/14/09
12/11/09
12/10/09
12/9/09
12/8/09
12/7/09
12/4/09
12/3/09
12/2/09
12/1/09
-0.06
11/30/09
-0.05
A
Appendix
Table A1. Sponsors,
competitors
and "endorsement
intensity."
Table
A.1: Sponsors,
competitors
and “endorsement intensity”
PROCTER & GAMBLE CO
Church & Dwight Co., Inc.
The Clorox Company
Colgate-Palmolive Company
Johnson & Johnson
CCA Industries, Inc.
Kimberly-Clark Corporation
Energizer Holdings, Inc.
Zep, Inc.
PC Group, Inc.
The Stephan Co.
NIKE INC
Deckers Outdoor Corp.
Crocs, Inc.
Skechers USA, Inc.
K-Swiss Inc.
Steven Madden, Ltd.
Heelys, Inc.
LaCrosse Footwear, Inc.
The Global Housing Group
adidas AG (ADR)
Puma AG Rudolf Dassler
PEPSICO INC
The Coca-Cola Company
Coca-Cola Enterprises (bottler)
Hansen Natural Corporation
Jones Soda Co. ( USA )
Cott Corporation (USA)
Dr Pepper Snapple Group
National Beverage Corp.
Reed's, Inc.
Celsius Holdings, Inc.
Fomento Economico Mexi
T L C VISION CORP
LCA-Vision Inc.
Hanger Orthopedic Grou
U.S. Physical Therapy,
NovaMed, Inc.
UCI Medical Affiliates
Pacific Health Care Or
Clinica de Marly S.A.
SHL TeleMedicine Ltd.
Feelgood Svenska AB
European Lifecare Grou
ACCENTURE LTD BERMUDA
Microsoft Corporation
Hewlett-Packard Company
Intl. Business Machine
Genpact Limited
Oracle Corporation
Infosys Tech. Ltd. (ADR)
Hewitt Associates, Inc.
Dell Inc.
Towers Watson & Co
Accenture Plc (Germany)
ELECTRONIC ARTS INC
THQ Inc.
Microsoft Corporation
Activision Blizzard, Inc.
Take-Two Interactive Software
The Walt Disney Company
KONAMI CORPORATION (ADR)
Sony Corporation (ADR)
Majesco Entertainment Co.
Time Warner Inc.
Nintendo Co., Ltd (ADR)
A T & T INC
Verizon Communications
Sprint Nextel Corporation
Qwest Communications I
CenturyTel, Inc.
Apple Inc.
General Communication,
Cbeyond, Inc.
Cincinnati Bell Inc.
Intl. Business Machine
Deutsche Telekom AG (ADR)
Notes: Each underlined heading is for one of the sponsors listed in Table 1. Next ten rows under each heading
show the first ten firms listed, in order, by Google Finance under "competitors." Competitors are measured
relative to the parent company. Bold competitors are those classified as "endorsement-intensive," meaning that
a Google search for the company name followed by "endorsement deals" yields at least one mention of a
celebrity endorsement contract. Competitor names in italics are not listed on U.S. stock exchanges.
40
Table A.2: Abnormal Returns and Alternative Measures of Search Intensity
Dependent variable: Sponsors' daily abnormal return
All Firms
Primary
Tiger Brand
"Tiger Woods endorsement" search intensity
-0.008*
(0.003)
-0.007
(0.004)
-0.026***
(0.006)
"Tiger Woods accident" search intensity
-0.003
(0.006)
0.000
(0.007)
-0.007
(0.010)
Constant
0.001
(0.002)
0.050
-0.001
(0.002)
0.056
0.007*
(0.003)
0.314
"Tiger Woods accident" search intensity
0.003
(0.005)
0.006
(0.006)
0.012
(0.011)
Constant
-0.002*
(0.001)
0.003
-0.004***
(0.001)
0.012
-0.005*
(0.002)
0.029
"Tiger Woods endorsement" search intensity
-0.007*
(0.003)
-0.007*
(0.004)
-0.024***
(0.006)
"Tiger Woods wife" search intensity
0.004
(0.003)
0.001
(0.004)
-0.002
(0.006)
Constant
-0.001
(0.002)
0.062
-0.001
(0.002)
0.057
0.006
(0.003)
0.308
"Tiger Woods wife" search intensity
0.003
(0.003)
0.001
(0.004)
-0.002
(0.007)
Constant
-0.003*
(0.001)
0.013
105
-0.004*
(0.002)
0.001
75
-0.003
(0.003)
0.003
45
R-squared
R-squared
R-squared
R-squared
Observations
Notes: Coefficients are from model (3) in text, modeling the relationship between firm-level daily
abnormal returns during the period [November 30, December 18] and alternative measures of
endorsement-related news intensity. Measures are the level of "Tiger Woods accident," "Tiger
Woods wife" and "Tiger Woods endorsement" search intensity on a [0, 1] scale from Figure 1.
Numbers in parentheses are standard errors. Asterisks indicate significance at 10% (*), 5%(**)
and 1%(***) or better.
41